RESOLVING THE LEDGER
This is one outlet's own report from PBS NewsHour — the article as it was filed. Other outlets are covering the same event; open the full story to compare every source side by side.
See the full story · 5 sources
Meta, the company behind Facebook and Instagram, has released a free artificial intelligence tool. The release was accompanied by a 14-page essay from CEO Mark Zuckerberg championing his view of how AI could transform our world and what protections are required. William Brangham discussed more with Mike Isaac of The New York Times. Geoff Bennett: Meta, the company behind Facebook and Instagram, has released a new free artificial intelligence tool today. The release was accompanied by a 14-page essay from Meta CEO Mark Zuckerberg championing his view of how artificial intelligence could transform our world and what protections are required. Our William Brangham as more. William Brangham: That's right, Geoff. Meta's new program is called Muse Glimmer. And it's what's known as an open-source model, meaning that anyone can download it, modify it, and use it however they want. In his essay, Zuckerberg makes the case that A.I. tools should all be like this, widely available and open, rather than walled off and controlled by a handful of powerful individuals or companies. He argues that the way forward is through -- quote -- "openness and putting the power of tech into more people's hands, not fewer." This puts Meta at odds with rivals like OpenAI and Anthropic, who argue that these increasingly powerful A.I. tools need to be tightly controlled because of the potential risks. So, for more on this growing debate, we are joined again by Mike Isaac, who's The New York Times' Silicon Valley correspondent. Mike, thank you so much for being here. Help us understand a little bit more. What do we mean when we say something is open-source versus the opposite? What's the distinction here? Mike Isaac, The New York Times: Yes, so open-source software has been around basically as long as Silicon Valley, but it's the idea that you're creating a computer program or coding language that can be sort of freely shared, copied, used, and manipulated to build your own types of programs, which is pretty different than companies like OpenAI and Anthropic, which build so-called closed-source, closed programs that people can't replicate. And the idea being the more people with access to the code, the more different types of programs or companies perhaps that you could build in the future. William Brangham: And it seems that Zuckerberg is arguing that it's more egalitarian, and that if you have more eyes on an open-source system, you will spot problems sooner, rather than later. But, as you mentioned, OpenAI and Anthropic, some of Meta's chief competitors, have an opposite view and think this stuff is too dangerous. What are they arguing as to why it should be closed? Mike Isaac: Yes, this debate goes for as long as computing has kind of been a thing. But it's exactly that. The folks of open-source say, hey, more eyeballs equals safer, we can figure out different security issues. The closed-source people, OpenAI and Anthropic, in particular, say, look, we love open-source software as much as anyone, but how A.I. is developing, how quickly, how powerful computers are becoming really is different than any sort of other computer program we have ever encountered. And so their argument is, a lot of this stuff, especially Anthropic, a lot of this stuff should stay closed, should not be replicable, and if used by the wrong folks or people who let's say don't have American interests in mind, for instance, they might be used to harm people on the planet. And so essentially what they're saying is, you should trust us and the American government to produce the right type of A.I. software, which clearly Mark Zuckerberg has taken issue with. William Brangham: I mean, I know we are early days with A.I., but is there any evidence one way or another that open-source systems are more inherently safe and less likely to fall into dangerous hands? Mike Isaac: I mean, I do think there are legitimate trade-offs and arguments on both sides to this. On the one hand, if anyone can see how a computer program performs or how it's operating, the idea being, like, the whole world can pick it apart faster and improve upon it, let's say. But it also, in terms of open-source programs, if anyone can use these things, and particularly very advanced software, if someone wants to cause havoc or do things that are really potentially disruptive to any different part of the economy -- or now there's talk of being able to develop biological weapons or computer programs that do hacking of different companies. That's really dangerous. Those are real risks that I think folks are wrestling through right now. And I think the American government, I will say, is trying to grapple with a balance between who should have control and how much control is too much control when developing these programs. So I honestly can see both sides of this. William Brangham: I mean, Zuckerberg is pitching this partly as a philosophical argument, but Meta and his competitors are all for-profit companies. There's an economic side to this as well. So -- and Meta's new tool comes as the Chinese have been putting out open-source inexpensive and quite effective A.I. tools. So, how much of this is philosophy and how much of this should we see this as competition? Mike Isaac: I think you nailed it. Everything that comes out of Mark Zuckerberg's, I guess, pen in this instance, is always a little bit self-serving, if not very much self-serving. Look, I think he actually does believe in this sort of egalitarian stance around code and who should have access to these tools. It's why he preaches everyone should be able to share on Facebook or Instagram or media should not be controlled just by a few outlets and everyone can sort of express themselves online. But I think, in this case, Meta is very behind in this A.I. race. They have spent billions of dollars developing data centers, they have hired top talent across the industry, and they're still getting beaten by Anthropic and OpenAI. So this ...
AIPROPX is an independent multi-source news index — we track, compare, and connect coverage from across the web into one place you won't find anywhere else.